Incompatible Geometry

[llpastro]

I have encountered an "Incompatible geometry" error message when using LinearFit.  I am trying to apply it to a synthetic lum and the extracted lum from an RGB image.  Both are in a nonlinear state.  What kinds of things could trigger this error?  Any suggestions will be appreciated.  I have tried redoing a dynamic crop.  I also tried to realign the two images but that failed, I assume, due to them being nonlinear and lots of processing applied.  Or is LinearFit not compatible with nonlinear images?

Larry

[sreilly]

Larry,

Why crop before combining? If the images are aligned,don't crop first, then take a aligned images into ImageIntergration and if you have over 25 images use Linear Fit Combine. What has been your approach so far? These are all the same filter? Actually I'm lost on what you are trying to do. Sounds like you have an all RGB image in which you extracted the luminace data as another image. Those two are already aligned and you would use a process such as LRGB combine to combine them. I'm lost with the use of Linear Fit at all at this point.

Steve

[RickS]

You'll get the incompatible geometry error if the images are different sizes or different types (e.g. one is mono and one is RGB.)

Cheers,
Rick.

[llpastro]

I was using  a workflow that I have picked up from one of the many online tutorials.  I don't remember which.  I am trying to create an LRGB image using an RGB image and a luminance image.  I had notes to use either LRGBCombination or using LinearFit.  I tried the LRGBCombination and got an RGB image that had no color.  So I tried the LinearFit method which has me extracting the luminance component from the RGB and then using the luminance image as reference, I applied it to the RGB luminance component which gave me the error message.  This precedes the use of ChannelCombination using LAB color space.

As regards the dyanamic cropping and aligning, all of the tutorials I have seen have you cropping as soon as you have each of the individual filter masters.  I used the same image for star alignment for all filters so they were already aligned.  I had no issues processing the RGB image.

Larry

[pfile]

LRGBCombination wants to run on stretched images, so you need to process your RGB and L separately. they need to have been stretched 'similarly' to one another; after that what you do is apply just the L image to the RGB image using LRGBCombination, essentially replacing the virtual L component of the RGB image with the L image.

the LinearFit before RGB combination could be unnecessary depending on your circumstances. but the idea behind it is to get the brightness of the 3 (linear) RGB channels about the same before doing ChannelCombination or PixelMath to create a linear RGB image. assuming that the first thing you do after RGB combine is DBE, as long as you run DBE with normalization turned off (the default) the channel brightness is going to end up being similar because the DBE essentially neutralizes the background. the only reason to do linear fit would be if one of the 3 images is particularly bright compared to the others, such that you have to crank the DBE sample tolerance way up to capture the background in that channel, thereby possibly oversampling the other channels. an alternate solution here is do just do DBE to the 3 master images separately before combining them.


rob

[sreilly]

Suggest taking the RGB image and the extracted luminance images all un-cropped and then combine. All the images are already aligned, at least I assume they are. Then with the combined image you can crop away. Are you using a color camera (CCD or dSLR), DeBayer process to get the R, G, B, images? I'd love to see this workflow to make and LRGB image using Linear Fit, can you provide a link to it please? BTW, the two processes I know of is the LRGB Combine and Color Combination for just the RGB and then using Dynamic Align if they are different in size. I bin my color 2x2 and luminance is binned 1x1 so that's how I combine them. RGB master in Color Combined and then Dynamic Align, and then LRGB having both images open and populating the L channel with the Lum and un-checking the R, G, and B boxes. Drag the triangle over to the luminance image.

Steve

[llpastro]

I will take a look at this again.  If I can find sources for my workflow, I will let you know.  Thanks for the responses.

Larry

[sreilly]

I hadn't noticed the ST-10XME camera in your signature but that being the case the workflow should be something along the lines of:

Calibrate all images with master darks and bias frames. (I save after each giving meaningful extensions to indicate their state such as m101-L-001_c, for calibarted, and so on)
Possible Cosmetic Correction if needed
Reject all bad images
Align all like binned images (if they are all 1x1, align all at one time and save again)
Use appropriate Image Integration method to produce master frames, save (m101-L-19, showing master L from 19 frames)
Use Color Combine or LRGB to make master RGB image, save again (ie: m101-rgb-121312, indicates 12 red, 13, green, and 12 blue frames used)
Process such as Dynamic Crop to get rid of the overlapping edges and save
Process the RGB master such as DBE, Masked Stretch or Histogram Stretch, noise and so on, save.
Process the master Lum as well, Dynamic Crop, DBE, stretch, noise, and so on,save
When happy with the 2 master frames, use Dynamic Align to align images. I use stars from all over the image, all corners, middle, and so on, in this case maybe 8-10 stars to align. Save each
Use LRGB combine to make LRGB image. uncheck all image boxes except the L, populate that image in the dialog, and then drag the triangle over the RGB image to combine.

This is all very basic and there are many processes that could be used in between depending on tastes and quality of data. That's where you develop your own style and routines.

HTH,

Steve

[llpastro]

Thanks again Steve.

Larry

[llpastro]

Steve, you mention Color Combine.  I do not find that as a process nor a script.  Do you mean ChannelCombination?

Larry

[sreilly]

Sorry, yes. I have a group of commonly used processes load on startup and named the process icons.

[llpastro]

Steve, I got the idea of using LinearFit before combining a Lum with an RGB from Warren Keller's Video tutorial, track PI3-3.

Larry

[Warhen]

Hey Larry, FWIW, I don't do this myself anymore. I prefer to 'eyeball' the chrominance stretch, just to the point where there is color contribution to dim areas, but no more. Since luminance will drive the lightness component of the combined images, I let it do that, and preserve saturation by not stretching the chrominance equal to the lum. Just another idea/approach.